Reply to Constantin-Teodosiu et al.: Mice with genetic PDH activation are not protected from high-fat diet–induced muscle insulin resistance

We thank Constantin-Teodosiu et al. (1) for their comments on our manuscript (2). Constantin-Teodosiu et al. interpret our data to indicate that mice with genetic activation of pyruvate dehydrogenase (PDH) by deletion of pyruvate dehydrogenase kinase isoforms 2 and 4 (DKO mice) are protected from high-fat diet (HFD)-induced muscle insulin resistance. Although Constantin-Teodosiu et al. correctly point out that HFD feeding did not worsen insulin-stimulated muscle glucose uptake in DKO mice (compare figure 2h and figure S4f from ref. 2), comparison with chow-fed WT control mice reveals that insulin-stimulated muscle glucose uptake was decreased by 50% in DKO mice on both regular chow and HFD (compare figure 2h and figure S4f of ref. 2). Therefore, to claim that DKO mice are protected from HFD-induced muscle insulin resistance is misleading. Such a claim overlooks the critical point that both chow-fed and HFD-fed DKO mice exhibit profound muscle insulin resistance. With respect to the therapeutic potential of pharmacological PDH activators, these data are therefore somewhat inauspicious.

Constantin-Teodosiu et al. also interpret our V_PDH/V_TCA flux measurements in DKO mice to indicate protection from HFD-induced muscle insulin resistance. However, our observation that V_PDH/V_TCA was similarly increased in DKO mice on regular chow (figure 1b of ref. 2) and an HFD (figure S4a of ref. 2) is consistent with the nature of our genetic model, which produced dephosphorylated, constitutively activated PDH that was insensitive to normal modifiers of PDH activity. Importantly, DKO mice displayed marked muscle insulin resistance despite large increases in muscle V_PDH/V_TCA flux compared with WT controls, indicating that glucose oxidation plays a relatively minor role in insulin-stimulated muscle glucose metabolism.

Constantin-Teodosiu et al. also suggest that the Randle glucose-fatty acid cycle, in which metabolites of fatty acid oxidation accumulate and inhibit glucose oxidation, is a viable model for lipid-induced insulin resistance in muscle. However, a substantial body of evidence indicates that the Randle hypothesis does not account for lipid-induced insulin resistance in human skeletal muscle (3). For example, a key prediction of the Randle hypothesis is that allosteric inhibition of phosphofructokinase-1 will cause buildup of its substrate, glucose-6-phosphate, and subsequent inhibition of hexokinase. However, intramyocellular glucose-6-phosphate and intramyocellular glucose concentrations are both decreased in humans with lipid-induced muscle insulin resistance, which is inconsistent with this hypothesis (4). Instead, these results imply that the defect in insulin-stimulated muscle glucose utilization under these conditions is secondary to impaired insulin-stimulated glucose transport activity, which will lead to blunted insulin stimulation of V_PDH/V_TCA flux as observed in the WT mice used in our study (compare figure 1b, black bars, and figure S4a, black bars, of ref. 2).

Finally, the PDH activators listed by Constantin-Teodosiu et al., exercise and dichloroacetate, exert protean physiological effects (5, 6) that render specific effects on relative PDH flux in skeletal muscle difficult to parse. Although the ultimate test of the therapeutic potential of PDH activators will of course be clinical trials of such agents, we maintain that the muscle insulin resistance observed in DKO mice fed a regular chow diet suggests that PDH activation will not be a viable therapeutic approach for type 2 diabetes.